Web lists-archives.com

Re: If Linux Is About Choice, Why Then ...




On Mon, Apr 10, 2017 at 11:13 PM, Nicolas George <george@xxxxxxxx> wrote:
> Le primidi 21 germinal, an CCXXV, tomas@xxxxxxxxxx a écrit :
>> SysV init is broken because it has no process monitoring? No.
>> Process monitoring isn't in its scope.
>
> Your other arguments make sense, but sorry, this one does not. The
> process with PID one is the only immortal process on the system, and
> adopts all orphan processes. For that reason, any kind of process
> monitoring, if it needs reliability, must be rooted in PID 1. And in
> turn, that makes process monitoring in scope for any project that aims
> to implement a program for PID 1.

No, it means you need to redesign certain aspects of the permissions systems.

systemd essentially does that through cgroups and something else I'm
not placing, but it shouldn't be at pid 1. It can be moved off to a
higher pid daemon. It takes refactoring that ripples into the kernel.
So does systemd.

The big problem with systemd was that the design was to put it all in
pid 1, which Linus rightly protested.

> And that is what makes SysV init unsalvageable.

... without refactoring. It takes work. We should quite fighting
against the idea that something takes work.

> Socket activation, automounting, etc., are entirely optional and
> peripheral. Process monitoring is not.

Socket activation is not optional. Automounting is. Process monitoring
may not be optional, but it has to be kept out of pid 1 because pid 1
has to be kept small.

> Regards,
>
> --
>   Nicolas George



-- 
Joel Rees

I'm imagining I'm a novelist:
http://joel-rees-economics.blogspot.com/2017/01/soc500-00-00-toc.html
More of my delusions:
http://reiisi.blogspot.jp/p/novels-i-am-writing.html